[
https://issues.apache.org/jira/browse/GOBBLIN-2052?focusedWorklogId=918425&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-918425
]
ASF GitHub Bot logged work on GOBBLIN-2052:
-------------------------------------------
Author: ASF GitHub Bot
Created on: 08/May/24 21:29
Start Date: 08/May/24 21:29
Worklog Time Spent: 10m
Work Description: ZihanLi58 commented on code in PR #3932:
URL: https://github.com/apache/gobblin/pull/3932#discussion_r1586589191
##########
gobblin-yarn/src/main/java/org/apache/gobblin/yarn/YarnAutoScalingManager.java:
##########
@@ -105,6 +114,10 @@ public class YarnAutoScalingManager extends
AbstractIdleService {
private final double overProvisionFactor;
private final SlidingWindowReservoir slidingFixedSizeWindow;
private static int maxIdleTimeInMinutesBeforeScalingDown =
DEFAULT_MAX_CONTAINER_IDLE_TIME_BEFORE_SCALING_DOWN_MINUTES;
+ private final int maxTimeInMinutesBeforeReleasingContainerHavingStuckTask;
+ private final boolean enableReleasingContainerHavingStuckTask;
+ private final boolean enableDetectStuckTask;
+ private final HashSet<TaskPartitionState> detectionForTaskStates;
Review Comment:
stucktaskStates?
##########
gobblin-yarn/src/main/java/org/apache/gobblin/yarn/YarnAutoScalingManager.java:
##########
@@ -136,6 +149,32 @@ public YarnAutoScalingManager(GobblinApplicationMaster
appMaster) {
DEFAULT_TASK_NUMBER_OF_ATTEMPTS_THRESHOLD);
this.splitWorkUnitReachThreshold = ConfigUtils.getBoolean(this.config,
SPLIT_WORKUNIT_REACH_ATTEMPTS_THRESHOLD,
DEFAULT_SPLIT_WORKUNIT_REACH_ATTEMPTS_THRESHOLD);
+ this.maxTimeInMinutesBeforeReleasingContainerHavingStuckTask =
ConfigUtils.getInt(this.config,
+ STUCK_TASK_CONTAINER_RELEASE_THRESHOLD_MINUTES,
+
DEFAULT_MAX_TIME_MINUTES_TO_RELEASE_CONTAINER_HAVING_HELIX_TASK_THAT_IS_STUCK);
+ this.enableReleasingContainerHavingStuckTask =
ConfigUtils.getBoolean(this.config,
+ RELEASE_CONTAINER_IF_TASK_IS_STUCK, false);
+ this.enableDetectStuckTask = ConfigUtils.getBoolean(this.config,
DETECT_IF_TASK_IS_STUCK, false);
+ this.detectionForTaskStates = getTaskStatesForWhichDetectionIsEnabled();
+ }
+
+ private HashSet<TaskPartitionState>
getTaskStatesForWhichDetectionIsEnabled() {
+ HashSet<TaskPartitionState> taskStates = new HashSet<>();
+ if (this.enableDetectStuckTask) {
+ List<String> taskStatesEnabledForDetection =
ConfigUtils.getStringList(this.config, ENABLE_DETECTION_FOR_TASK_STATES);
+ for (String taskState : taskStatesEnabledForDetection) {
+ try {
+ taskStates.add(TaskPartitionState.valueOf(taskState));
+ } catch (IllegalArgumentException e) {
+ log.warn("Invalid task state {} provided for detection, ignoring",
taskState);
+ }
+ }
+ if (taskStatesEnabledForDetection.isEmpty()) {
+ // if config was not set, default case enable only for INIT state
+ taskStates.add(TaskPartitionState.INIT);
Review Comment:
Can we set default config value to init instead of adding it here
separately?
Issue Time Tracking
-------------------
Worklog Id: (was: 918425)
Time Spent: 2h (was: 1h 50m)
> Release container which is running yarn task that is stuck in INIT state
> ------------------------------------------------------------------------
>
> Key: GOBBLIN-2052
> URL: https://issues.apache.org/jira/browse/GOBBLIN-2052
> Project: Apache Gobblin
> Issue Type: Improvement
> Components: gobblin-yarn
> Reporter: pradeep pallikila
> Assignee: Abhishek Tiwari
> Priority: Major
> Time Spent: 2h
> Remaining Estimate: 0h
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
